1 /* ------------------------------------------------------------------ 2 * Copyright (C) 1998-2009 PacketVideo 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either 13 * express or implied. 14 * See the License for the specific language governing permissions 15 * and limitations under the License. 16 * ------------------------------------------------------------------- 17 */ 18 19 #ifndef _MP4LIB_INT_H_ 20 #define _MP4LIB_INT_H_ 21 22 #include "mp4def.h" 23 #include "mp4enc_api.h" 24 #include "rate_control.h" 25 26 /* BitstreamEncVideo will be modified */ 27 typedef struct tagBitstream 28 { 29 Int(*writeVideoPacket)(UChar *buf, Int nbytes_required); /*write video packet out */ 30 UChar *bitstreamBuffer; /*buffer to hold one video packet*/ 31 Int bufferSize; /*total bitstream buffer size in bytes */ 32 Int byteCount; /*how many bytes already encoded*/ 33 UInt word; /*hold one word temporarily */ 34 Int bitLeft; /*number of bits left in "word" */ 35 UChar* overrunBuffer; /* pointer to overrun buffer */ 36 Int oBSize; /* length of overrun buffer */ 37 struct tagVideoEncData *video; 38 } BitstreamEncVideo; 39 40 typedef struct tagVOP 41 { 42 PIXEL *allChan; /* [yuv]Chan point into this buffer */ 43 PIXEL *yChan; /* The Y component */ 44 PIXEL *uChan; /* The U component */ 45 PIXEL *vChan; /* The V component */ 46 Int frame; /* frame number */ 47 Int volID; /* Layer number */ 48 //Int timeStamp; /* Vop TimeStamp in msec */ 49 50 /* Syntax elements copied from VOL (standard) */ 51 Int width; /* Width (multiple of 16) */ 52 Int height; /* Height (multiple of 16) */ 53 Int pitch; /* Pitch (differs from width for UMV case) */ 54 Int padded; /* flag whether this frame has been padded */ 55 56 /* Actual syntax elements for VOP (standard) */ 57 Int predictionType; /* VOP prediction type */ 58 Int timeInc; /* VOP time increment (relative to last mtb) */ 59 Int vopCoded; 60 Int roundingType; 61 Int intraDCVlcThr; 62 Int quantizer; /* VOP quantizer */ 63 Int fcodeForward; /* VOP dynamic range of motion vectors */ 64 Int fcodeBackward; /* VOP dynamic range of motion vectors */ 65 Int refSelectCode; /* enhancement layer reference select code */ 66 67 /* H.263 parameters */ 68 Int gobNumber; 69 Int gobFrameID; 70 Int temporalRef; /* temporal reference, roll over at 256 */ 71 Int temporalInterval; /* increase every 256 temporalRef */ 72 73 } Vop; 74 75 typedef struct tagVol 76 { 77 Int volID; /* VOL identifier (for tracking) */ 78 Int shortVideoHeader; /* shortVideoHeader mode */ 79 Int GOVStart; /* Insert GOV Header */ 80 Int timeIncrementResolution; /* VOL time increment */ 81 Int nbitsTimeIncRes; /* number of bits for time increment */ 82 Int timeIncrement; /* time increment */ 83 Int moduloTimeBase; /* internal decoder clock */ 84 Int prevModuloTimeBase; /* in case of pre-frameskip */ 85 86 Int fixedVopRate; 87 BitstreamEncVideo *stream; /* library bitstream buffer (input buffer) */ 88 89 /* VOL Dimensions */ 90 Int width; /* Width */ 91 Int height; /* Height */ 92 93 /* Error Resilience Flags */ 94 Int ResyncMarkerDisable; /* VOL Disable Resynch Markers */ 95 Int useReverseVLC; /* VOL reversible VLCs */ 96 Int dataPartitioning; /* VOL data partitioning */ 97 98 /* Quantization related parameters */ 99 Int quantPrecision; /* Quantizer precision */ 100 Int quantType; /* MPEG-4 or H.263 Quantization Type */ 101 102 /* Added loaded quant mat, 05/22/2000 */ 103 Int loadIntraQuantMat; /* Load intra quantization matrix */ 104 Int loadNonIntraQuantMat; /* Load nonintra quantization matrix */ 105 Int iqmat[64]; /* Intra quant.matrix */ 106 Int niqmat[64]; /* Non-intra quant.matrix */ 107 108 109 /* Parameters used for scalability */ 110 Int scalability; /* VOL scalability (flag) */ 111 Int scalType; /* temporal = 0, spatial = 1, both = 2 */ 112 113 Int refVolID; /* VOL id of reference VOL */ 114 Int refSampDir; /* VOL resol. of ref. VOL */ 115 Int horSamp_n; /* VOL hor. resampling of ref. VOL given by */ 116 Int horSamp_m; /* sampfac = hor_samp_n/hor_samp_m */ 117 Int verSamp_n; /* VOL ver. resampling of ref. VOL given by */ 118 Int verSamp_m; /* sampfac = ver_samp_n/ver_samp_m */ 119 Int enhancementType; /* VOL type of enhancement layer */ 120 121 /* These variables were added since they are used a lot. */ 122 Int nMBPerRow, nMBPerCol; /* number of MBs in each row & column */ 123 Int nTotalMB; 124 Int nBitsForMBID; /* how many bits required for MB number? */ 125 126 /* for short video header */ 127 Int nMBinGOB; /* number of MBs in GOB, 05/22/00 */ 128 Int nGOBinVop; /* number of GOB in Vop 05/22/00 */ 129 } Vol; 130 131 typedef struct tagMacroBlock 132 { 133 Int mb_x; /* X coordinate */ 134 Int mb_y; /* Y coordinate */ 135 Short block[9][64]; /* 4-Y, U and V blocks , and AAN Scale*/ 136 } MacroBlock; 137 138 typedef struct tagRunLevelBlock 139 { 140 Int run[64]; /* Runlength */ 141 Int level[64]; /* Abs(level) */ 142 Int s[64]; /* sign level */ 143 } RunLevelBlock; 144 145 typedef struct tagHeaderInfoDecVideo 146 { 147 UChar *Mode; /* Modes INTRA/INTER/etc. */ 148 UChar *CBP; /* MCBPC/CBPY stuff */ 149 } HeaderInfoEncVideo; 150 151 typedef Short typeDCStore[6]; /* ACDC */ 152 typedef Short typeDCACStore[4][8]; 153 154 typedef struct tagMOT 155 { 156 Int x; /* half-pel resolution x component */ 157 Int y; /* half-pel resolution y component */ 158 Int sad; /* SAD */ 159 } MOT; 160 161 typedef struct tagHintTrackInfo 162 { 163 UChar MTB; 164 UChar LayerID; 165 UChar CodeType; 166 UChar RefSelCode; 167 168 } HintTrackInfo; 169 170 171 typedef struct tagVideoEncParams 172 { 173 //Int Width; /* Input Width */ 174 //Int Height; /* Input Height */ 175 //float FrameRate; /* Input Frame Rate */ 176 UInt TimeIncrementRes; /* timeIncrementRes */ 177 178 /*VOL Parameters */ 179 Int nLayers; 180 Int LayerWidth[4]; /* Encoded Width */ 181 Int LayerHeight[4]; /* Encoded Height */ 182 float LayerFrameRate[4]; /* Encoded Frame Rate */ 183 Int LayerBitRate[4]; /* Encoded BitRate */ 184 Int LayerMaxBitRate[4]; /* Maximum Encoded BitRate */ 185 float LayerMaxFrameRate[4]; /* Maximum Encoded Frame Rate */ 186 Int LayerMaxMbsPerSec[4]; /* Maximum mbs per second, according to the specified profile and level */ 187 Int LayerMaxBufferSize[4]; /* Maximum buffer size, according to the specified profile and level */ 188 189 Bool ResyncMarkerDisable; /* Disable Resync Marker */ 190 Bool DataPartitioning; /* Base Layer Data Partitioning */ 191 Bool ReversibleVLC; /* RVLC when Data Partitioning */ 192 Bool ACDCPrediction; /* AC/DC Prediction */ 193 Int QuantType[4]; /* H263, MPEG2 */ 194 Int InitQuantBvop[4]; 195 Int InitQuantPvop[4]; 196 Int InitQuantIvop[4]; 197 Int ResyncPacketsize; 198 199 Int RoundingType; 200 Int IntraDCVlcThr; 201 202 /* Rate Control Parameters */ 203 MP4RateControlType RC_Type; /*Constant Q, M4 constantRate, VM5+, M4RC,MPEG2TM5 */ 204 205 /* Intra Refresh Parameters */ 206 Int IntraPeriod; /* Intra update period */ 207 Int Refresh; /* Number of MBs refresh in each frame */ 208 /* Other Parameters */ 209 Bool SceneChange_Det; /* scene change detection */ 210 Bool FineFrameSkip_Enabled; /* src rate resolution frame skipping */ 211 Bool VBR_Enabled; /* VBR rate control */ 212 Bool NoFrameSkip_Enabled; /* do not allow frame skip */ 213 Bool NoPreSkip_Enabled; /* do not allow pre-skip */ 214 215 Bool H263_Enabled; /* H263 Short Header */ 216 Bool GOV_Enabled; /* GOV Header Enabled */ 217 Bool SequenceStartCode; /* This probably should be removed */ 218 Bool FullSearch_Enabled; /* full-pel exhaustive search motion estimation */ 219 Bool HalfPel_Enabled; /* Turn Halfpel ME on or off */ 220 Bool MV8x8_Enabled; /* Enable 8x8 motion vectors */ 221 Bool RD_opt_Enabled; /* Enable operational R-D optimization */ 222 Int GOB_Header_Interval; /* Enable encoding GOB header in H263_WITH_ERR_RES and SHORT_HERDER_WITH_ERR_RES */ 223 Int SearchRange; /* Search range for 16x16 motion vector */ 224 Int MemoryUsage; /* Amount of memory allocated */ 225 Int GetVolHeader[2]; /* Flag to check if Vol Header has been retrieved */ 226 Int BufferSize[2]; /* Buffer Size for Base and Enhance Layers */ 227 Int ProfileLevel[2]; /* Profile and Level for encoding purposes */ 228 float VBV_delay; /* VBV buffer size in the form of delay */ 229 Int maxFrameSize; /* maximum frame size(bits) for H263/Short header mode, k*16384 */ 230 Int profile_table_index; /* index for profile and level tables given the specified profile and level */ 231 232 } VideoEncParams; 233 234 /* platform dependent functions */ 235 typedef struct tagFuncPtr 236 { 237 // Int (*SAD_MB_HalfPel)(UChar *ref,UChar *blk,Int dmin_lx,Int xh,Int yh,void *extra_info); 238 Int(*SAD_MB_HalfPel[4])(UChar*, UChar*, Int, void *); 239 Int(*SAD_Blk_HalfPel)(UChar *ref, UChar *blk, Int dmin, Int lx, Int rx, Int xh, Int yh, void *extra_info); 240 Int(*SAD_Macroblock)(UChar *ref, UChar *blk, Int dmin_lx, void *extra_info); 241 Int(*SAD_Block)(UChar *ref, UChar *blk, Int dmin, Int lx, void *extra_info); 242 Int(*SAD_MB_PADDING)(UChar *ref, UChar *blk, Int dmin, Int lx, void *extra_info); /*, 4/21/01 */ 243 void (*ComputeMBSum)(UChar *cur, Int lx, MOT *mot_mb); 244 void (*ChooseMode)(UChar *Mode, UChar *cur, Int lx, Int min_SAD); 245 void (*GetHalfPelMBRegion)(UChar *cand, UChar *hmem, Int lx); 246 void (*blockIdct)(Int *block); 247 248 249 } FuncPtr; 250 251 /* 04/09/01, for multipass rate control */ 252 253 typedef struct tagRDInfo 254 { 255 Int QP; 256 Int actual_bits; 257 float mad; 258 float R_D; 259 } RDInfo; 260 261 typedef struct tagMultiPass 262 { 263 /* multipass rate control data */ 264 Int target_bits; /* target bits for current frame, = rc->T */ 265 Int actual_bits; /* actual bits for current frame obtained after encoding, = rc->Rc*/ 266 Int QP; /* quantization level for current frame, = rc->Qc*/ 267 Int prev_QP; /* quantization level for previous frame */ 268 Int prev_prev_QP; /* quantization level for previous frame before last*/ 269 float mad; /* mad for current frame, = video->avgMAD*/ 270 Int bitrate; /* bitrate for current frame */ 271 float framerate; /* framerate for current frame*/ 272 273 Int nRe_Quantized; /* control variable for multipass encoding, */ 274 /* 0 : first pass */ 275 /* 1 : intermediate pass(quantization and VLC loop only) */ 276 /* 2 : final pass(de-quantization, idct, etc) */ 277 /* 3 : macroblock level rate control */ 278 279 Int encoded_frames; /* counter for all encoded frames */ 280 Int re_encoded_frames; /* counter for all multipass encoded frames*/ 281 Int re_encoded_times; /* counter for all times of multipass frame encoding */ 282 283 /* Multiple frame prediction*/ 284 RDInfo **pRDSamples; /* pRDSamples[30][32], 30->30fps, 32 -> 5 bit quantizer, 32 candidates*/ 285 Int framePos; /* specific position in previous multiple frames*/ 286 Int frameRange; /* number of overall previous multiple frames */ 287 Int samplesPerFrame[30]; /* number of samples per frame, 30->30fps */ 288 289 /* Bit allocation for scene change frames and high motion frames */ 290 float sum_mad; 291 Int counter_BTsrc; /* BT = Bit Transfer, bit transfer from low motion frames or less complicatedly compressed frames */ 292 Int counter_BTdst; /* BT = Bit Transfer, bit transfer to scene change frames or high motion frames or more complicatedly compressed frames */ 293 float sum_QP; 294 Int diff_counter; /* diff_counter = -diff_counter_BTdst, or diff_counter_BTsrc */ 295 296 /* For target bitrate or framerate update */ 297 float target_bits_per_frame; /* = C = bitrate/framerate */ 298 float target_bits_per_frame_prev; /* previous C */ 299 float aver_mad; /* so-far average mad could replace sum_mad */ 300 float aver_mad_prev; /* previous average mad */ 301 Int overlapped_win_size; /* transition period of time */ 302 Int encoded_frames_prev; /* previous encoded_frames */ 303 } MultiPass; 304 305 /* End */ 306 307 #ifdef HTFM 308 typedef struct tagHTFM_Stat 309 { 310 Int abs_dif_mad_avg; 311 UInt countbreak; 312 Int offsetArray[16]; 313 Int offsetRef[16]; 314 } HTFM_Stat; 315 #endif 316 317 /* Global structure that can be passed around */ 318 typedef struct tagVideoEncData 319 { 320 /* VOL Header Initialization */ 321 UChar volInitialize[4]; /* Used to Write VOL Headers */ 322 /* Data For Layers (Scalability) */ 323 Int numberOfLayers; /* Number of Layers */ 324 Vol **vol; /* Data stored for each VOL */ 325 326 /* Data used for encoding frames */ 327 VideoEncFrameIO *input; /* original input frame */ 328 Vop *currVop; /* Current reconstructed VOP */ 329 Vop *prevBaseVop; /* Previous reference Base Vop */ 330 Vop *nextBaseVop; /* Next reference Base Vop */ 331 Vop *prevEnhanceVop;/* Previous Enhancement Layer Vop */ 332 Vop *forwardRefVop; /* Forward Reference VOP */ 333 Vop *backwardRefVop; /* Backward Reference VOP */ 334 335 /* scratch memory */ 336 BitstreamEncVideo *bitstream1; /* Used for data partitioning */ 337 BitstreamEncVideo *bitstream2; /* and combined modes as */ 338 BitstreamEncVideo *bitstream3; /* intermediate storages */ 339 340 UChar *overrunBuffer; /* extra output buffer to prevent current skip due to output buffer overrun*/ 341 Int oBSize; /* size of allocated overrun buffer */ 342 343 Int dc_scalar_1; /*dc scalar for Y block */ 344 Int dc_scalar_2; /*dc scalar for U, V block*/ 345 346 /* Annex L Rate Control */ 347 rateControl *rc[4]; /* Pointer to Rate Control structure*/ 348 /* 12/25/00, each R.C. for each layer */ 349 350 /********* motion compensation related variables ****************/ 351 MOT **mot; /* Motion vectors */ 352 /* where [mbnum][0] = 1MV. 353 [mbnum][1...4] = 4MVs 354 [mbnum][5] = backward MV. 355 [mbnum][6] = delta MV for direct mode. 356 [mbnum][7] = nothing yet. */ 357 UChar *intraArray; /* Intra Update Arrary */ 358 float sumMAD; /* SAD/MAD for frame */ 359 360 /* to speedup the SAD calculation */ 361 void *sad_extra_info; 362 #ifdef HTFM 363 Int nrmlz_th[48]; /* Threshold for fast SAD calculation using HTFM */ 364 HTFM_Stat htfm_stat; /* For statistics collection */ 365 #endif 366 367 /*Tao 04/09/00 For DCT routine */ 368 UChar currYMB[256]; /* interleaved current macroblock in HTFM order */ 369 MacroBlock *outputMB; /* Output MB to VLC encode */ 370 UChar predictedMB[384]; /* scrath memory for predicted value */ 371 RunLevelBlock RLB[6]; /* Run and Level of coefficients! */ 372 Short dataBlock[128]; /* DCT block data before and after quant/dequant*/ 373 374 UChar bitmaprow[8]; /* Need to keep it for ACDCPrediction, 8 bytes for alignment, need only 6 */ 375 UChar bitmapcol[6][8]; 376 UInt bitmapzz[6][2]; /* for zigzag bitmap */ 377 Int zeroMV; /* flag for zero MV */ 378 379 Int usePrevQP; /* flag for intraDCVlcThreshold switch decision */ 380 Int QP_prev; /* use for DQUANT calculation */ 381 Int *acPredFlag; /* */ 382 typeDCStore *predDC; /* The DC coeffs for each MB */ 383 typeDCACStore *predDCAC_row; 384 typeDCACStore *predDCAC_col; 385 386 387 UChar *sliceNo; /* Slice Number for each MB */ 388 389 Int header_bits; /* header bits in frmae */ 390 HeaderInfoEncVideo headerInfo; /* MB Header information */ 391 UChar zz_direction; /* direction of zigzag scan */ 392 UChar *QPMB; /* Quantizer value for each MB */ 393 394 /* Miscellaneous data points to be passed */ 395 float FrameRate; /* Src frame Rate */ 396 397 ULong nextModTime; /* expected next frame time */ 398 UInt prevFrameNum[4]; /* previous frame number starting from modTimeRef */ 399 UInt modTimeRef; /* Reference modTime update every I-Vop*/ 400 UInt refTick[4]; /* second aligned referenc tick */ 401 Int relLayerCodeTime[4];/* Next coding time for each Layer relative to highest layer */ 402 403 ULong modTime; /* Input frame modTime */ 404 Int currLayer; /* Current frame layer */ 405 Int mbnum; /* Macroblock number */ 406 407 /* slice coding, state variables */ 408 Vop *tempForwRefVop; 409 Int tempRefSelCode; 410 Int end_of_buf; /* end of bitstream buffer flag */ 411 Int slice_coding; /* flag for slice based coding */ 412 Int totalSAD; /* So far total SAD for a frame */ 413 Int numIntra; /* So far number of Intra MB */ 414 Int offset; /* So far MB offset */ 415 Int ind_x, ind_y; /* So far MB coordinate */ 416 Int collect; 417 Int hp_guess; 418 /*********************************/ 419 420 HintTrackInfo hintTrackInfo; /* hintTrackInfo */ 421 /* IntraPeriod, Timestamp, etc. */ 422 float nextEncIVop; /* counter til the next I-Vop */ 423 float numVopsInGOP; /* value at the beginning of nextEncIVop */ 424 425 /* platform dependent functions */ 426 FuncPtr *functionPointer; /* structure containing platform dependent functions */ 427 428 /* Application controls */ 429 VideoEncControls *videoEncControls; 430 VideoEncParams *encParams; 431 432 MultiPass *pMP[4]; /* for multipass encoding, 4 represents 4 layer encoding */ 433 434 } VideoEncData; 435 436 /*************************************************************/ 437 /* VLC structures */ 438 /*************************************************************/ 439 440 typedef struct tagVLCtable 441 { 442 unsigned int code; /* right justified */ 443 int len; 444 } VLCtable, *LPVLCtable; 445 446 447 /*************************************************************/ 448 /* Approx DCT */ 449 /*************************************************************/ 450 typedef struct struct_approxDCT approxDCT; 451 struct struct_approxDCT 452 { 453 Void(*BlockDCT8x8)(Int *, Int *, UChar *, UChar *, Int, Int); 454 Void(*BlockDCT8x8Intra)(Int *, Int *, UChar *, UChar *, Int, Int); 455 Void(*BlockDCT8x8wSub)(Int *, Int *, UChar *, UChar *, Int, Int); 456 }; 457 458 /*************************************************************/ 459 /* QP structure */ 460 /*************************************************************/ 461 462 struct QPstruct 463 { 464 Int QPx2 ; 465 Int QP; 466 Int QPdiv2; 467 Int QPx2plus; 468 Int Addition; 469 }; 470 471 472 #endif /* _MP4LIB_INT_H_ */ 473 474
